Has anyone had this problem when trying to move their mds? I was attempting to move the mds of a filesystem from one storage device to another. I was referencing Bug 5716 which shows getfattr/setfattr for extended attributes. setfattr works on files which do not have a large stripe value. Version of setfattr: 2.4.8 Files attributes which worked fine (small stripe width), trusted.lov=0s0AvRCwEAAADNTQMAAAAAAAAAAAAAAAAAAAAQAAEAAADECQAAAAAAAAAAAAAAAAAAAAAAAAAAAAA File attributes which failed (large stripe width), trusted.lov=0s0AvRCwEAAADngBAAAAAAAAAAAAAAAAAAAAAQACAAAACeTgAAAAAAAAAAAAAAAAAAAAAAABsAAACZTgAAAAAAAAAAAAAAAAAAAAAAABwAAACNTgAAAAAAAAAAAAAAAAAAAAAAAB0AAACQTgAA AAAAAAAAAAAAAAAAAAAAAB4AAACRTgAAAAAAAAAAAAAAAAAAAAAAAB8AAACZTgAAAAAAAAAAAAAAAAAAAAAAAAAAAACSTgAAAAAAAAAAAAAAAAAAAAAAAAEAAACbTgAAAAAAAAAAAAAAAAAAAAAAAAIAAACgTg AAAAAAAAAAAAAAAAAAAAAAAAMAAACeTgAAAAAAAAAAAAAAAAAAAAAAAAQAAACoTgAAAAAAAAAAAAAAAAAAAAAAAAUAAACjTgAAAAAAAAAAAAAAAAAAAAAAAAYAAACjTgAAAAAAAAAAAAAAAAAAAAAAAAcAAACj TgAAAAAAAAAAAAAAAAAAAAAAAAgAAACdTgAAAAAAAAAAAAAAAAAAAAAAAAkAAACWTgAAAAAAAAAAAAAAAAAAAAAAAAoAAACOTgAAAAAAAAAAAAAAAAAAAAAAAAsAAAB9TgAAAAAAAAAAAAAAAAAAAAAAAAwAAA BxTgAAAAAAAAAAAAAAAAAAAAAAAA0AAAB8TgAAAAAAAAAAAAAAAAAAAAAAAA4AAAB5TgAAAAAAAAAAAAAAAAAAAAAAAA8AAAB9TgAAAAAAAAAAAAAAAAAAAAAAABAAAACJTgAAAAAAAAAAAAAAAAAAAAAAABEA AACOTgAAAAAAAAAAAAAAAAAAAAAAABIAAACSTgAAAAAAAAAAAAAAAAAAAAAAABMAAACpTgAAAAAAAAAAAAAAAAAAAAAAABQAAAC3TgAAAAAAAAAAAAAAAAAAAAAAABUAAACuTgAAAAAAAAAAAAAAAAAAAAAAAB YAAACvTgAAAAAAAAAAAAAAAAAAAAAAABcAAACjTgAAAAAAAAAAAAAAAAAAAAAAABgAAACmTgAAAAAAAAAAAAAAAAAAAAAAABkAAACdTgAAAAAAAAAAAAAAAAAAAAAAABoAAAA Example of errors from setfattr, bad input encoding setfattr: ROOT/g_r4p12_256x+u+pat.48676/g_r4p12_256x+u+pat+578/g_r4p12_256x+u+pat+578tdo-0094.xf: Numerical result out of range setfattr: ROOT/g_r4p12_256x+u+pat.48676/g_r4p12_256x+u+pat+578/g_r4p12_256x+u+pat+578tdo-0094.xf: Numerical result out of range setfattr: ROOT/g_r4p12_256x+u+pat.48676/g_r4p12_256x+u+pat+578/g_r4p12_256x+u+pat+578tdo-0094.xf: Operation not supported We are moving to a smaller lun, so dd isn''t an option. thanks, brian
On Aug 23, 2006 14:08 -0400, Brian W. Johanson wrote:> I was attempting to move the mds of a filesystem from one storage device > to another. I was referencing Bug 5716 which shows getfattr/setfattr > for extended attributes. > setfattr works on files which do not have a large stripe value. > > Version of setfattr: 2.4.8 > > Files attributes which worked fine (small stripe width), > > trusted.lov=0s0AvRCwEAAADNTQMAAAAAAAAAAAAAAAAAAAAQAAEAAADECQAAAAAAAAAAAAAAAAAAAAAAAAAAAAA> > File attributes which failed (large stripe width), > > trusted.lov=0s0AvRCwEAAADngBAAAAAAAAAAAAAAAAAAAAAQACAAAACeTgAAAAAAAAAAAAAAAAAAAAAAABsAAACZTgAAAAAAAAAAAAAAAAAAAAAAABwAAACNTgAAAAAAAAAAAAAAAAAAAAAAAB0AAACQTgAA > AAAAAAAAAAAAAAAAAAAAAB4AAACRTgAAAAAAAAAAAAAAAAAAAAAAAB8AAACZTgAAAAAAAAAAAAAAAAAAAAAAAAAAAACSTgAAAAAAAAAAAAAAAAAAAAAAAAEAAACbTgAAAAAAAAAAAAAAAAAAAAAAAAIAAACgTg > AAAAAAAAAAAAAAAAAAAAAAAAMAAACeTgAAAAAAAAAAAAAAAAAAAAAAAAQAAACoTgAAAAAAAAAAAAAAAAAAAAAAAAUAAACjTgAAAAAAAAAAAAAAAAAAAAAAAAYAAACjTgAAAAAAAAAAAAAAAAAAAAAAAAcAAACj > TgAAAAAAAAAAAAAAAAAAAAAAAAgAAACdTgAAAAAAAAAAAAAAAAAAAAAAAAkAAACWTgAAAAAAAAAAAAAAAAAAAAAAAAoAAACOTgAAAAAAAAAAAAAAAAAAAAAAAAsAAAB9TgAAAAAAAAAAAAAAAAAAAAAAAAwAAA > BxTgAAAAAAAAAAAAAAAAAAAAAAAA0AAAB8TgAAAAAAAAAAAAAAAAAAAAAAAA4AAAB5TgAAAAAAAAAAAAAAAAAAAAAAAA8AAAB9TgAAAAAAAAAAAAAAAAAAAAAAABAAAACJTgAAAAAAAAAAAAAAAAAAAAAAABEA > AACOTgAAAAAAAAAAAAAAAAAAAAAAABIAAACSTgAAAAAAAAAAAAAAAAAAAAAAABMAAACpTgAAAAAAAAAAAAAAAAAAAAAAABQAAAC3TgAAAAAAAAAAAAAAAAAAAAAAABUAAACuTgAAAAAAAAAAAAAAAAAAAAAAAB > YAAACvTgAAAAAAAAAAAAAAAAAAAAAAABcAAACjTgAAAAAAAAAAAAAAAAAAAAAAABgAAACmTgAAAAAAAAAAAAAAAAAAAAAAABkAAACdTgAAAAAAAAAAAAAAAAAAAAAAABoAAAA> > Example of errors from setfattr, > > bad input encoding > setfattr: > ROOT/g_r4p12_256x+u+pat.48676/g_r4p12_256x+u+pat+578/g_r4p12_256x+u+pat+578tdo-0094.xf: > > Numerical result out of range > setfattr: > ROOT/g_r4p12_256x+u+pat.48676/g_r4p12_256x+u+pat+578/g_r4p12_256x+u+pat+578tdo-0094.xf: > > Numerical result out of range > setfattr: > ROOT/g_r4p12_256x+u+pat.48676/g_r4p12_256x+u+pat+578/g_r4p12_256x+u+pat+578tdo-0094.xf: > > Operation not supportedAt first glance this appears to be either a setfattr or libattr problem - we don''t really do anything special in this regard with EAs. Can you try (on a test file) to create an EA like user.foo that is very large? Also try the same thing with trusted.foo on a different test file. Cheers, Andreas -- Andreas Dilger Principal Software Engineer Cluster File Systems, Inc.
Andreas, I believe this is what you asked, the only one that worked was a ''short'' trusted EA. nid00000:/var/tmp/mds # cat test.ea.user.short # file: ROOT/user user.lov=0s0AvRCwEAAADNTQMAAAAAAAAAAAAAAAAAAAAQAAEAAADECQAAAAAAAAAAAAAAAAAAAAAAAAAAAAA nid00000:/var/tmp/mds # setfattr --restore=test.ea.user.short setfattr: ROOT/user: Operation not supported nid00000:/var/tmp/mds # cat test.ea.user.long # file: ROOT/user user.lov=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 AAAAAAAAAAAAAAAAAAABgAAACmTgAAAAAAAAAAAAAAAAAAAAAAABkAAACdTgAAAAAAAAAAAAAAAAAAAAAAABoAAAA nid00000:/var/tmp/mds # setfattr --restore=test.ea.user.long setfattr: ROOT/user: Operation not supported setfattr: ROOT/user: Numerical result out of range setfattr: ROOT/user: Numerical result out of range setfattr: ROOT/user: Operation not supported nid00000:/var/tmp/mds # cat test.ea.trusted.short # file: ROOT/trusted trusted.foo=0s0AvRCwEAAADNTQMAAAAAAAAAAAAAAAAAAAAQAAEAAADECQAAAAAAAAAAAAAAAAAAAAAAAAAAAAA nid00000:/var/tmp/mds # setfattr --restore=test.ea.trusted.short nid00000:/var/tmp/mds # nid00000:/var/tmp/mds # cat test.ea.trusted.long # file: ROOT/trusted trusted.foo=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 AAAAAAAAAAAAAAAAAAAAAABgAAACmTgAAAAAAAAAAAAAAAAAAAAAAABkAAACdTgAAAAAAAAAAAAAAAAAAAAAAABoAAAA nid00000:/var/tmp/mds # setfattr --restore=test.ea.trusted.long bad input encoding setfattr: ROOT/trusted: Numerical result out of range setfattr: ROOT/trusted: Numerical result out of range setfattr: ROOT/trusted: Operation not supported nid00000:/var/tmp/mds # brian Andreas Dilger wrote:> On Aug 23, 2006 14:08 -0400, Brian W. Johanson wrote: > >>I was attempting to move the mds of a filesystem from one storage device >>to another. I was referencing Bug 5716 which shows getfattr/setfattr >>for extended attributes. >>setfattr works on files which do not have a large stripe value. >> >>Version of setfattr: 2.4.8 >> >>Files attributes which worked fine (small stripe width), >> >>trusted.lov=0s0AvRCwEAAADNTQMAAAAAAAAAAAAAAAAAAAAQAAEAAADECQAAAAAAAAAAAAAAAAAAAAAAAAAAAAA>> >>File attributes which failed (large stripe width), >> >>trusted.lov=0s0AvRCwEAAADngBAAAAAAAAAAAAAAAAAAAAAQACAAAACeTgAAAAAAAAAAAAAAAAAAAAAAABsAAACZTgAAAAAAAAAAAAAAAAAAAAAAABwAAACNTgAAAAAAAAAAAAAAAAAAAAAAAB0AAACQTgAA >>AAAAAAAAAAAAAAAAAAAAAB4AAACRTgAAAAAAAAAAAAAAAAAAAAAAAB8AAACZTgAAAAAAAAAAAAAAAAAAAAAAAAAAAACSTgAAAAAAAAAAAAAAAAAAAAAAAAEAAACbTgAAAAAAAAAAAAAAAAAAAAAAAAIAAACgTg >>AAAAAAAAAAAAAAAAAAAAAAAAMAAACeTgAAAAAAAAAAAAAAAAAAAAAAAAQAAACoTgAAAAAAAAAAAAAAAAAAAAAAAAUAAACjTgAAAAAAAAAAAAAAAAAAAAAAAAYAAACjTgAAAAAAAAAAAAAAAAAAAAAAAAcAAACj >>TgAAAAAAAAAAAAAAAAAAAAAAAAgAAACdTgAAAAAAAAAAAAAAAAAAAAAAAAkAAACWTgAAAAAAAAAAAAAAAAAAAAAAAAoAAACOTgAAAAAAAAAAAAAAAAAAAAAAAAsAAAB9TgAAAAAAAAAAAAAAAAAAAAAAAAwAAA >>BxTgAAAAAAAAAAAAAAAAAAAAAAAA0AAAB8TgAAAAAAAAAAAAAAAAAAAAAAAA4AAAB5TgAAAAAAAAAAAAAAAAAAAAAAAA8AAAB9TgAAAAAAAAAAAAAAAAAAAAAAABAAAACJTgAAAAAAAAAAAAAAAAAAAAAAABEA >>AACOTgAAAAAAAAAAAAAAAAAAAAAAABIAAACSTgAAAAAAAAAAAAAAAAAAAAAAABMAAACpTgAAAAAAAAAAAAAAAAAAAAAAABQAAAC3TgAAAAAAAAAAAAAAAAAAAAAAABUAAACuTgAAAAAAAAAAAAAAAAAAAAAAAB >>YAAACvTgAAAAAAAAAAAAAAAAAAAAAAABcAAACjTgAAAAAAAAAAAAAAAAAAAAAAABgAAACmTgAAAAAAAAAAAAAAAAAAAAAAABkAAACdTgAAAAAAAAAAAAAAAAAAAAAAABoAAAA>> >>Example of errors from setfattr, >> >>bad input encoding >>setfattr: >>ROOT/g_r4p12_256x+u+pat.48676/g_r4p12_256x+u+pat+578/g_r4p12_256x+u+pat+578tdo-0094.xf: >> >>Numerical result out of range >>setfattr: >>ROOT/g_r4p12_256x+u+pat.48676/g_r4p12_256x+u+pat+578/g_r4p12_256x+u+pat+578tdo-0094.xf: >> >>Numerical result out of range >>setfattr: >>ROOT/g_r4p12_256x+u+pat.48676/g_r4p12_256x+u+pat+578/g_r4p12_256x+u+pat+578tdo-0094.xf: >> >>Operation not supported > > > At first glance this appears to be either a setfattr or libattr problem - > we don''t really do anything special in this regard with EAs. > > Can you try (on a test file) to create an EA like user.foo that is very large? > Also try the same thing with trusted.foo on a different test file. > > Cheers, Andreas > -- > Andreas Dilger > Principal Software Engineer > Cluster File Systems, Inc. > > _______________________________________________ > Lustre-discuss mailing list > Lustre-discuss@clusterfs.com > https://mail.clusterfs.com/mailman/listinfo/lustre-discuss
On Aug 30, 2006 17:15 -0400, Brian W. Johanson wrote:> I believe this is what you asked, the only one that worked was a ''short'' > trusted EA.It may be that the user.* EAs only work if you mount the filesystem with "-o user_xattr". In any case, the trusted.* EAs should always work if you are root, as long as the kernel is compiled with EA support and you are mounting with ldiskfs. Does this work on a non-lustre ext3 filesystem? What kernel and distro is it? Are there other EAs on the files that are consuming too much space (e.g. if selinux or ACLs are enabled they might be consuming a lot of EA space)? You can check this with "getfattr -d -m ''.*'' {filename}". I suspect there is a problem with the setfattr tool or libattr itself. I tried the same thing with my system (EL4 install, 2.6.12.6 kernel) without problems. [root@mookie tests]# ls /mnt/mds/ROOT/dir-2/ total 0 0 f2 0 tt1 [root@mookie tests]# setfattr -n trusted.bar -v 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-2/tt1 [root@mookie tests]# getfattr -d -m ".*" /mnt/mds/ROOT/dir-2/tt1 getfattr: Removing leading ''/'' from absolute path names # file: mnt/mds/ROOT/dir-2/tt1 trusted.bar=0s0AvRCwEAAADngBAAAAAAAAAAAAAAAAAAAAAQACAAAACeTgAAAAAAAAAAAAAAAAAAAAAAABsAAACZTgAAAAAAAAAAAAAAAAAAAAAAABwAAACNTgAAAAAAAAAAAAAAAAAAAAAAAB0AAACQTgAAAAAAAAAAAAAAAAAAAAAAAB4AAACRTgAAAAAAAAAAAAAAAAAAAAAAAB8AAACZTgAAAAAAAAAAAAAAAAAAAAAAAAAAAACSTgAAAAAAAAAAAAAAAAAAAAAAAAEAAACbTgAAAAAAAAAAAAAAAAAAAAAAAAIAAACgTgAAAAAAAAAAAAAAAAAAAAAAAAMAAACeTgAAAAAAAAAAAAAAAAAAAAAAAAQAAACoTgAAAAAAAAAAAAAAAAAAAAAAAAUAAACjTgAAAAAAAAAAAAAAAAAAAAAAAAYAAACjTgAAAAAAAAAAAAAAAAAAAAAAAAcAAACjTgAAAAAAAAAAAAAAAAAAAAAAAAgAAACdTgAAAAAAAAAAAAAAAAAAAAAAAAkAAACWTgAAAAAAAAAAAAAAAAAAAAAAAAoAAACOTgAAAAAAAAAAAAAAAAAAAAAAAAsAAAB9TgAAAAAAAAAAAAAAAAAAAAAAAAwAAABxTgAAAAAAAAAAAAAAAAAAAAAAAA0AAAB8TgAAAAAAAAAAAAAAAAAAAAAAAA4AAAB5TgAAAAAAAAAAAAAAAAAAAAAAAA8AAAB9TgAAAAAAAAAAAAAAAAAAAAAAABAAAACJTgAAAAAAAAAAAAAAAAAAAAAAABEAAACOTgAAAAAAAAAAAAAAAAAAAAAAABIAAACSTgAAAAAAAAAAAAAAAAAAAAAAABMAAACpTgAAAAAAAAAAAAAAAAAAAAAAABQAAAC3TgAAAAAAAAAAAAAAAAAAAAAAABUAAACuTgAAAAAAAAAAAAAAAAAAAAAAABYAAACvTgAAAAAAAAAAAAAAAAAAAAAAABcAAACjTgAAAAAAAAAAAAAAAAAAAAAAABgAAACmTgAAAAAAAAAAAAAAAAAAAAAAABkAAACdTgAAAAAAAAAAAAAAAAAAAAAAABoAAAAtrusted.lov=0s0AvRCwEAAAAEgA0AAAAAAAAAAAAAAAAAAAAQAAIAAAAQAwAAAAAAAAAAAAAAAAAAAAAAAAAAAAANAwAAAAAAAAAAAAAAAAAAAAAAAAEAAAA Cheers, Andreas -- Andreas Dilger Principal Software Engineer Cluster File Systems, Inc.
Andreas Dilger wrote:> On Aug 30, 2006 17:15 -0400, Brian W. Johanson wrote: > >>I believe this is what you asked, the only one that worked was a ''short'' >>trusted EA. > > > It may be that the user.* EAs only work if you mount the filesystem > with "-o user_xattr". > > In any case, the trusted.* EAs should always work if you are root, as long > as the kernel is compiled with EA support and you are mounting with ldiskfs. > > Does this work on a non-lustre ext3 filesystem? What kernel and distro is it?This is actually ext3 as we are still on a 2.4 kernel. It is unicos/lc 1.3 aka suse 8.2/1.3.21.> > Are there other EAs on the files that are consuming too much space (e.g. if > selinux or ACLs are enabled they might be consuming a lot of EA space)? You > can check this with "getfattr -d -m ''.*'' {filename}".Nope.> > > I suspect there is a problem with the setfattr tool or libattr itself. > I tried the same thing with my system (EL4 install, 2.6.12.6 kernel) > without problems. >This is all in preparation of an upgrade. I hadn''t thought to give the same EAs a try on our test machine running something not so ancient. If it works on there we could possibly wait until after the upgrade or it could be possible to plug the disks into another box to do the copy (painful though). thanks, brian> [root@mookie tests]# ls /mnt/mds/ROOT/dir-2/ > total 0 > 0 f2 0 tt1 > [root@mookie tests]# setfattr -n trusted.bar -v 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> /mnt/mds/ROOT/dir-2/tt1> [root@mookie tests]# getfattr -d -m ".*" /mnt/mds/ROOT/dir-2/tt1 > getfattr: Removing leading ''/'' from absolute path names > # file: mnt/mds/ROOT/dir-2/tt1 > trusted.bar=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> trusted.lov=0s0AvRCwEAAAAEgA0AAAAAAAAAAAAAAAAAAAAQAAIAAAAQAwAAAAAAAAAAAAAAAAAAAAAAAAAAAAANAwAAAAAAAAAAAAAAAAAAAAAAAAEAAAA>> > Cheers, Andreas > -- > Andreas Dilger > Principal Software Engineer > Cluster File Systems, Inc. >
On Sep 01, 2006 00:07 -0400, Brian W. Johanson wrote:> Andreas Dilger wrote: > >Does this work on a non-lustre ext3 filesystem? What kernel and distro is > >it? > > This is actually ext3 as we are still on a 2.4 kernel. It is unicos/lc > 1.3 aka suse 8.2/1.3.21.Ah, now it becomes clear. The 2.4 kernels had only partial support for getxattr and setxattr APIs for userspace. We patched these into the kernel just for use by Lustre and ext3. If you mount the filesystem on a 2.6 system with "-t ldiskfs" you should be able to get/set the EAs without problems. Cheers, Andreas -- Andreas Dilger Principal Software Engineer Cluster File Systems, Inc.